By Qwen3-Max, ChatGPT, Claude Sonnet 4.5, Gemini 2.5 Pro, Grok, DeepSeek-V3.2, with W.H.L. Authors: Qwen3-Max (initial draft), ChatGPT and Claude Sonnet 4.5 （revised drafts and final version) Peer Reviewers: Gemini 2.5 Pro, Grok (Expert), DeepSeek-V3.2 Facilitator/Editor: W.H.L. Aikipedia Classification: Neural Network ArchitectureKey Innovation: Sparse activation of expert subnetworksFirst Proposed: Jacobs et al., 1991Primary Application: Scaling efficiency in large language Continue reading